OmniParser V2 introduces SPOT prompting to create a single model for four visual text parsing tasks, reports competitive results on eight datasets, and shows the prompting works inside multimodal LLMs.
Textocr: Towards large-scale end-to-end reasoning for arbitrary- shaped scene text,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
OmniParser V2: Structured-Points-of-Thought for Unified Visual Text Parsing and Its Generality to Multimodal Large Language Models
OmniParser V2 introduces SPOT prompting to create a single model for four visual text parsing tasks, reports competitive results on eight datasets, and shows the prompting works inside multimodal LLMs.